From: Christoph Neubauer <christoph.neubauer@...>
Date: 2004-09-15T21:20:05+09:00
Subject: Re: How to Control Time Behaviour of System Call ?

Hi Ara, Robert !

All attempts to control runtime behaviour of system call 'from outside'
failed.
Threads, Pipes, ... on ruby level everything was started/killed fine,
but on system level, the activated command remainded alive until its natural
end.

So we decided to implement a different solution for our time control.

We integrate a watchdog mechanism into one of our 'runtime libraries'.
When a test case is started, a watchdog is started in background, too.
When a testcase ends, it generates a normalized minute about its status
(passed, failed, aborted, ...) and some other stuff.
If a testcase is blocked and/or the watchdog finds out, that there is a
timeout error,
the watchdog generates a 'timeout' minute and exits immediately, this way
aborting the testcase.

So we get a uniform behaviour/interface of our testcases,
regardless if they are started under supervision of a tester,
or via the regression script.

Many thanks for help,
Chris


"Ara.T.Howard" <Ara.T.Howard@noaa.gov> wrote in message
news:Pine.LNX.4.60.0409130931040.12299@harp.ngdc.noaa.gov...
> On Mon, 13 Sep 2004, Neubauer Christoph wrote:
>
> > Hi Ara !
> >
> > Many thanks for the info below.
> > Unfortunately I'm not in the position to use extra packages
> > (Sorry, I should have mentioned this in the posting),
> > and my Ruby machine answers:
> >
> > ./fork_timeout.rb:7:in `fork': The fork() function is unimplemented on
this
> > machine (NotImplementedError)
> > from ./fork_timeout.rb:7:in `timeout'
> > from ...
> >
> > Right at time I try to implement Robert's suggestion
> > of popen and parallel Threads, but I got stuck, somehow.
>
> perhaps something like this is what robert was suggesting, try some
windows
> specific commands with it:
>
> ~ > cat popen_timeout.rb
> module PopenTimeout
> #{{{
>    class TimeoutError < StandardError; end
>
>    def timeout n_sec, cmd
> #{{{
>      t =
>        Thread.new(cmd) do |cmd|
>          pipe = IO::popen(cmd)
>          stdout = pipe.read
>          pipe.close
>          status = $?.exitstatus
>          [stdout, status]
>        end
>
>      sleep n_sec
>
>      case t.status
>        when NilClass
>          raise t.value
>        when FalseClass
>          return t.value
>        else
>          t.kill
>          raise TimeoutError, "<#{ n_sec }>"
>      end
> #}}}
>    end
>    module_function 'timeout'
>    public 'timeout'
> #}}}
> end
>
> if $0 == __FILE__
>    stdout, status =
>      PopenTimeout::timeout(2, 'sleep 1 && echo 42')
>
>    puts stdout
>    puts status
>
>    stdout, status =
>      PopenTimeout::timeout(1, 'sleep 2 && echo 42')
> end
>
> > Maybe I have to find a completely different solution,
> > e.g. passing control of time behaviour into called functionality.
> > So it will return at all means, either regularly or 'killing' itself.
> >
> > Many thanks (I could learn a lot) and best regards from Vienna
> > Chris
> >
> > PS: What (kind of) organization is "noaa.gov" ?
>
>
> national oceananic and atmouspheric administration, check out
>
>    http://www.ngdc.noaa.gov/stp/stp.html
>
> to see what i do.  we work with satelite imagery showing the nighttime
lights
> of earth from space like
>
>    http://dmsp.ngdc.noaa.gov/html/night_light_posters.html
>
>
> > PPS: I like 'your' Dogen quotation !
>
> thanks - it's a constant reminder to myself to take things in stride.
>
> you are in germany?  what's the programming market like there?  we are
> thinking of moving!  ;-)
>
> -a
> --
>
============================================================================
===
> | EMAIL   :: Ara [dot] T [dot] Howard [at] noaa [dot] gov
> | PHONE   :: 303.497.6469
> | A flower falls, even though we love it;
> | and a weed grows, even though we do not love it.
> |   --Dogen
>
============================================================================
===